Nonlinear Systems

Mohammed Rida MOKHTARI

École Supérieure en Sciences Appliquées de Tlemcen-ESSAT

August 23, 2022

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 1 / 33
 Chapter 1

Chapter 1

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 2 / 33
 Chapter 1

Introduction to theory: basic definitions

A general nonlinear dynamic system may be modeled by a finite number of
coupled first-order ordinary differential equations.
ẋ = f (t, x, u) (1)
where x = [x1 , · · · , xn ]T ∈ Rn : state vector, n: system order,
u = [u1 , · · · , um ]T ∈ Rm : input vector (control, disturbance).
f(.) is a vector field in Rn : a function associating a vector to n-dim point x.
Initial condition: x(t0 ) = (x1 (0), x2 (t0 ), · · · , xn (t0 ))
Eq. (1) is called state equation.
Solution is in the form x(t0 , t) that defines a family of time trajectories in
the state space (also referred as phase space).Imposing the initial condition
x(t0 ) determines one unique trajectory.
Another equation named output equation:
y = h(t, x, u) (2)
where y = [y1 , · · · , yp ] ∈ R : output vector.
T p

Eqs. (1) and (2) together are called state-space model. . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 3 / 33
 Chapter 1

Introduction to theory: basic definitions

Analysis : unforced state equations

ẋ = f (x), f : Rn → Rn time-invariant (autonomous)

ẋ = f (t, x), f : R × Rn → Rn time-varying (Non-autonomous)

Control design:
ẋ = f (x, u), f : R n × Rm → R n autonomous with inputs
ẋ = f (t, x, u), f : R × Rn × Rm → Rn Non-autonomous with inputs
ẋ = f (x) + g(x)u, f : R n → Rn , g : R m → Rn affine in u

Linear systems: if f and h are linear functions of x and u

ẋ = f (x) → ẋ = Ax autonomous
ẋ = f (x, u) → ẋ = Ax + Bu

ẋ = f (t, x, u) → ẋ = A(t)x + B(t)u

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 4 / 33
 Chapter 1

Introduction to theory: basic definitions

For linear systems

The Linear Systems satisfy the superposition principals:

1 Homogeneity: f (αx) = αf (x), ∀α ∈ R.
2 Additivity: f (x + y) = f (x) + f (y), ∀x, y ∈ Rn .
Unique equilibrium point.
Stability is independent of initial conditions x(t) = x(0)eAt

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 5 / 33
 Chapter 1

Introduction to theory: basic definitions

For nonlinear systems

Non superposition principle =⇒ More complex behavior

Example : Under-water vehicle v̇ + |v|v = u

Settles faster in response to positive step.
Scaling input does not result into the same scaling in output.

u=1
⇒ 0 + |vs |vs = 1
⇒ vs = 1

u = 10
⇒ 0 + |vs |vs = √
10
⇒ vs = 10
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 6 / 33
 Chapter 1

Nonlinear System Examples

Figure: A nonlinear system is a system in which the change of the output is not
proportional to the change of the input.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 7 / 33
 Chapter 1

Nonlinear System Examples

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 8 / 33
 Chapter 1

Nonlinear System Examples

Systems with essential nonlinearities in the model

Unmanned Aerial vehicle model

ξ˙ = v
mv̇ = Tz Rη e3 + mge3 + Fext
η̇ = Qη Ω
J Ω̇ = −Ω × JΩ + τ + Mext

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 9 / 33
 Chapter 1

Nonlinear System Examples

Systems with saturation

ẋ = Ax + Bsat(u)
u = P ID(x) y


u If |u| ≤ 1
sat(u) =
sgn(u) If |u| ≥ 1
x
The output is proportional to input for a
limited range.
Output becomes constant if input is outside
this range.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 10 / 33
 Chapter 1

Common Nonlinearities
memoryless
y y
y
m

-d -d
x x x
d d

-m
y
y y
k(a-d) m
m
a>d
h h
x -a -d x −2 2 x
d a

-m
-m
Relay . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 11 / 33
 Chapter 1

Equilibrium points:

An equilibrium point represents a stationary condition of a dynamical sys-

tem. The state xe ∈ Rn is a fixed point for ẋ = f (x) if f (xe ) = 0, ∀t ≥ 0.

If a dynamical system has an initial condition x(0) at an equilibrium point xe ,

then it will stay at xe forever, i.e. x(t) = xe , ∀t ≥ 0.
Example 1: ẋ = x(x − 2)2
 xe = 0
f (x̄) = 0 ⇒ x(x − 2) = 0 ⇒
2
xe = 2
This system has two isolated equilibrium points at 0 and 2.
Example 2:
ẋ = sin(x), x(0) = x0 ∈ R (3)
x(t) ∈ R 1st order system (scalar state)

f (x̄) = sin(x̄) = 0 =⇒ x̄ = kπ, k = 0, ±1, ±2, · · ·

This system has infinite many equilibrium points. .
.
.
.
.
. . . . .
. . . .
. . . .
. . . .
. . . .
. . . . .
.
.
.
.
.
.
.
.
.

R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 12 / 33

 Chapter 1

Equilibrium points:
Non-isolated equilibria

For a linear system ẋ = Ax, A ∈ Rn×n .

If A is nonsingular (det(A)̸= 0), then x∗ = 0 is the unique equilibrium.
If A is singular (det(A)= 0), then Null space defines a continuum of
equilibria.

y
Example:
D = (-0.25, 1.25(
     )B = (1, 1.5

ẋ1 −1 −2 x1
=
ẋ2 −2 −4 x2 x
| {z } | {z } | {z }
ẋ A x

−2x1 − 4x2 = 0 )C = (-0.3, -1.25
ẋ = 0 ⇒ Ax = 0 ⇒
−x1 − 2x2 = 0
⇒ x1 = −2x2 )A = (-1.5, -2.5
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 13 / 33
 Chapter 1 AAA

Linear approximations around an equilibrium

A linear approximation of a system around an equilibrium point can be used to
study the behavior of a nonlinear system around the equilibrium point.

Local stability properties of xe can be determined by linearizing the vector

field f (x) at xe :

δf
f (xe + x̃) = f (xe ) + x̃ + higher order terms in (x − xe ) (4)
| {z } δx x=x̄
=0 | {z }
∆
=A

Thus, the linearized model is :

x˜˙ = Ax̃ (5)
At
The solution of the linearized form (5) has the form : x̃ = e x0
• If Re(λi (A)) < 0, then xe is locally asymptotically stable.
• If Re(λi (A)) > 0 for some eigenvalue λi of A, then x̄ is unstable.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 14 / 33
 Chapter 1 AAA

Linear approximations around an equilibrium

Example 1: Linearization of ẋ = sin(x)

∂f cos(2Kπ) = 1, n even
= cos(x)|x̄ =
∂x x̄ sin((2K + 1)π) = −1, n odd

x̄ = nπ, n = 0, ±1, ±2, · · · .

Example 2: (2nd order system)
   
f1 (x) x2
ẋ = =
f2 (x) −x1 + x21 x2 − x2
 
∂f1 (x) ∂f1 (x)
 
∂f  ∂x1 ∂x2
 0 1
= = −1 + 2x1 x2 x21 − 1
∂x ∂f1 (x) ∂f2 (x)
∂x1 ∂x2
Linearization around xe (0, 0) : δf
δx xe(x − xe)
  
0 1 x1
=
−1 −1 x2 . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 15 / 33
 Chapter 1 AAA

Caveats:
1 Only local properties can be determined from the linearization.
2 If Re(λi ) ≤ 0 with some e-values having Re(λi ) = 0, then
linearization is inconclusive as a stability test. Higher order terms
determine stability.

Example: (a) : ẋ = x3 , (b) : ẋ = −x3 .

In both cases, linearized systems around x̄ = 0 are the same: ẋ = 0 ⇒ x(t) = x0 ,
but NL systems have different behaviors.
Case (a) Case (b)
1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0

−0.2 −0.2

−0.4 −0.4

−0.6 −0.6

−0.8 −0.8

−1 −1
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1
. . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 16 / 33
 Essentially Nonlinear Phenomena

Essentially Nonlinear Phenomena

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 17 / 33
 Essentially Nonlinear Phenomena

1. Multiple Isolated equilibrium

Linear systems have only one equilibrium point at origin or a number of non
isolated equilibrium points.

Example: Pendulum (two isolated equilibia)

lmθ̈ = −klθ̇ − mg sin(θ) (6)
 
θ
Define x = . State space : S 1 × R
θ̇

ẋ1 = x2
(7)
ẋ2 = −m k
x2 − g
l sin x1

Equilibria: Two isolated E.P (0,0) and (π, 0).

Linearization :
  
   

0 1
(stable) at x1 = 0
δf 0 1  − gl − kl
= = (8)
δx − gl cos x1 − kl 
 0 1
 (unstable) at x1 = π
g
l − k
l . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 18 / 33
 Essentially Nonlinear Phenomena

Stability may depend upon initial conditions

Example : The logistic equation (population dynamics in isolation)

ẋ = f (x) = rx(k − x) x, r > 0, K > 0 (9)

| {z }
growth rate

x ∈ R+ is the population size and K is the carrying capacity.

Stability can be determined from the sign of f (x) around the equilibrium.
x(t)
(ẋe = 0) 40
Equilibrium points are : 
⇒ rxe (K − xe ) = 0 ⇒ xe = 0
xe = K 30
Linearization :
20
∂f
∂x = (rK − 2rx)|xe
 xe
x̄ = 0 → Kr → unstable. (10) 10
x̄ = K → −Kr → stable
!
t
x=0 ! =K 0 1 . . . 2. . . . . . . . 3. . . . . . 4. . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 19 / 33
 Essentially Nonlinear Phenomena

2. Finite Escape Time

In linear case :
Solution
ẋ = λx −→ x(t) = expλt x(0).

If λ > 0 =⇒ limt→∞ |x(t)| = +∞. Only as t → ∞, |x(t)| → ∞.

x(t)
In nonlinear case :
Example : ẋ = x2 , x(0) = x0 , x ∈ R
R x(t) Rt
dx
dt = x2 ⇒ x0 dx x2 = 0 dt
⇒ − x(t)
1
+ x10 = t − 0

1
x(t) = 1
x0 −t
x0
t
x0 > 0 ⇒ t → 1
x0 ⇒ x(t) → ∞ t

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 20 / 33
 Essentially Nonlinear Phenomena

3. Limit cycles
Linear oscillators exhibit a continuum of periodic orbits (closed orbit) :
x(t + T ) = x(t), ∀t > 0, for somme T > 0
Every circle is periodic orbit for ẋ = Ax where
 
0 −β
A= , (λ1,2 = ±jβ)
β 0
2 2

1 1

0 0

-1 -1

-2
-2
-2 -1 0 1 2
-2 -1 0 1 2

Every Periodic Orbit is a Cycle but not a Limit Cycle.

In contrast, a limit cycle is an isolated closed trajectory (closed orbit) and can
occur only in nonlinear systems. .
.
.
.
.
. . . . .
. . . .
. . . .
. . . .
. . . .
. . . . .
.
.
.
.
.
.
.
.
.

R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 21 / 33

 Essentially Nonlinear Phenomena

3. Limit cycles
Example: Harmonic oscillator

mẍ
|{z} + kx
|{z} =0 (11)
inertial term stiffness term

    
ẋ1 0 1 x1
=
ẋ2 −mk
0 x2
q
ω0 = k
m. (λ1,2 = ±jω0 ).
Amplitude of oscillations depends on initial conditions.
Can be destroyed by small modelling imperfections.

The harmonic oscillator has closed orbits but no limit cycles. Limit
cycles cannot be generated by LTI systems.
The linear oscillator is not structurally stable. A stable oscillators
must be produced by nonlinear systems.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 22 / 33
 Essentially Nonlinear Phenomena

3. Limit cycles
3

Example : Van der Pol oscillator 1

x, dx/dt
0

ẍ − µ(1 − x2 )ẋ + x = 0 −1

−2
If µ = 0 ⇒ ẍ + x = 0 ← simple harmonic −3
0 0.5 1 1.5 2 2.5 3
oscillator t

 Response with x(0) = 0.05, x˙ (0) = 0.05

ẋ1 = x2 4
(12)
ẋ2 = µ(1 − x21 )x2 − x1 3

Equilibrium point :ẋ = 0 ⇒ x̄ = [0 0]T . 1

dx/dt
0
Linarization around x̄ = (0, 0) 6 6 ?
−1
  −2
0 1
A= −3
−1 µ −4
−3 −2 −1 0 1 2 3
x

Positive sign of µ tells us that e.p is unstable. State trajectories x(t), ẋ(t)


. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 23 / 33
 Essentially Nonlinear Phenomena

3. Limit cycles: Examples of limit cycles

Stable Limite Cycle Unstable Limite Cycle SemiStable Limite Cycle

2 2 2

1 1 1

0 0 0

-1 -1 -1

-2 -2 -2

-2 -1 0 1 2 -2 -1 0 1 2 -2 -1 0 1 2

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 24 / 33
 Essentially Nonlinear Phenomena

4. Chaos :

Chaos
Irregular oscillations, never exactly repeating.
Behavior of nonlinear systems may be extremely sensive to small changes in
initial conditions/input/parameters.

Example : Lorenz system (attractor) derrived by Ed Lorenz in 1963 as a

simplified model of convection rolls in the atmosphere.

60

The lorenz system is a 3rd order 40

system (3 states x,y,z). z

20

ẋ = σ(y − x) −20
0 40
ẏ = rx − y − xz (13) −10 20

ż = xy − bz 0 0
x 10 −20 y
−40 20
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 25 / 33
 Essentially Nonlinear Phenomena

4. Chaos:
For continuous-time, time invariant systems, n ≥ 3 state variable required for
chaos.
No simple characterization of asymptotic behavior.
Huge sensitivity to initial conditions.
blue: (x, y, z) = (0, 1, 1.05 )
Choatic behavior with σ = 10, b=8/3, r=28
red: (x, y, z) = (0, 1, 1.050001)
20

x −20
0 5 10 15 20 25 30 35 40

50

0
y

−50
0 5 10 15 20 25 30 35 40

60

40
z

20

0
0 5 10 15
. . .
20
. . . .
25
. . . .
30
. . . .
35
. . .
40
. .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 26 / 33
 Essentially Nonlinear Phenomena

4. Chaos :
Example: The double pendulum (System is implicit for l1 ̸= l2 )


 (m1 + m1 )l1 θ̈1 + m2 l2 θ̈2 cos(θ1 − θ2 ) + m2 l2 θ̇2 sin(θ1 − θ2 )

+ g(m1 + m2 ) sin θ1 = 0

 m2 l2 θ̈2 + m2 l1 θ̈1 cos(θ1 − θ2 ) − m2 l1 θ̇12 sin(θ1 − θ2 )

+ m2 g sin θ2 = 0

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 27 / 33
 Essentially Nonlinear Phenomena

5. Bifurcation: Fold bifurcation

A bifurcation is an abrupt change in qualitative behavior as a parameter is varied.
Examples : creation (or death) of equilibrium points (or limit cycles) and/or
change of their stability properties.
Fold bifurcation: 1st order system
Example : ẋ = µ − x2 ,
Equilibrium
 √points :

 ± µ µ > 0 one stable equilibrium and one unstable equilibrium

x̄ = 0 µ = 0 single equilibrium (called a saddle)


none µ < 0 no equilibria

ẋ ẋ ẋ

x x x

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 28 / 33
 Essentially Nonlinear Phenomena

5. Bifurcation: Fold bifurcation

µc = 0 is the critical value of parameter µ which represents boundary be-

tween "no equilibrium points" and the presence of equilibrium points.

Creation/destruction of fixed points is called saddle node bifurcation

bifurcation diagram
Linearization :
 √
δf 2 µ unstable
= 2x̄ = √ x
δx x̄
−2 µ stable

µ > 0.
µ
Note:
Ac = δf
δx = 0 → linearization
x̄c =x̄( µc )
disappears, no information about stability of the
. . . . . . . . . . . . . . . . . . . .
system. . . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 29 / 33
 Essentially Nonlinear Phenomena

5. Bifurcation: Transcritical bifurcation

Transcritical bifurcation: bifurcation diagram
Example : ẋ = µx − x2 , x(t) ∈ R x
Equilibrium points : x̄ = 0. x̄ = µ.
Linearization : 
δf µ if x̄ = 0 µ
= µ − 2x̄ =
δx −µ if x̄ = µ

µ < 0 : x̄ = 0 is stable, x̄ = µ is unstable.

µ > 0 : x̄ = 0 is unstable, x̄ = µ is stable.
ẋ ẋ ẋ

x x x

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 30 / 33
 Essentially Nonlinear Phenomena

5. Bifurcation : Pichfork bifurcation

Pitchfork Bifurcation- 2 types : supercritical Pitchfork and subcritical
pitchfork
Example 1:
ẋ = µx − x3 supercritical

0
Equilibrium points : f (x̄) = 0 ⇒ x̄(µ − x̄2 ) = 0 ⇒ x̄ = √
± µ, µ>0

ẋ ẋ ẋ

x x x

2 equilibrium points emerge when we increase µ. . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 31 / 33
 Essentially Nonlinear Phenomena

5. Bifurcation : Pitchfork bifurcation

bifurcation diagram
x

“supercritical pitchfork"

Example 2: ẋ = µx + x3 , subcritical pitchfork.

Equilibrium points :

0
f (x̄) = 0 ⇒ x̄(µ − x̄ ) = 0 ⇒ x̄ =
2 √
± −µ, µ<0

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 32 / 33
 Essentially Nonlinear Phenomena

5.Bifurcation: Pitchfork bifurcation

Linearization :
δf
δx = µ
x̄=0
δf
δx √ = −2µ
x̄=± −µ

“subcritical pitchfork"

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 23, 2022 33 / 33